Chlorophyll Content Inversion Of Pepper Leaves Based On Hyperspectral Remote Sensing Dat

Pepper is an important economic crop in China,and it is also listed as one of the12 characteristic and advantageous industries in Guizhou Province.Therefore,doing a good job in pepper planting and production and promoting the development of pepper industry is an effective starting point for rural revitalization and enriching people’s income in Guizhou.Monitoring the growth status of pepper is an important prerequisite for the above work,and the accurate acquisition of chlorophyll content（SPAD value）in chili pepper leaves is the key.In this study,four kinds of pepper were used as the research object,and the SPAD values of pepper leaves were collected in the field,the near-ground hyperspectral data and the low-altitude hyperspectral data of UAV were obtained.The spectral reflectance of pepper leaves was extracted by hyperspectral data,and the data were preprocessed and analyzed.The genetic partial least squares algorithm（GA-PLS）was used to optimize the spectral bands,and the vegetation index with high correlation with pepper chlorophyll was constructed.The correlation between chlorophyll and hyperspectral data of pepper was analyzed,and a hyperspectral inversion model of chlorophyll with strong universality and high precision suitable for different data sources and different pepper varieties was constructed.The specific research contents and results are as follows:（1）The best inversion model of SPAD value of pepper leaves based on near-ground hyperspectral data was selected.From the correlation analysis and model construction based on the selected pepper sample SPAD value and near-ground hyperspectral data,it is found that the larger the chlorophyll value,the lower the reflectance of the leaves,which was inversely proportional.the spectral differentiation technology can effectively highlight the spectral information of pepper canopy.The SPAD value of pepper leaves had the highest correlation coefficient with the second-order differential spectrum.The accuracy of the model established by the Pearson-GAPLS dimension reduction method was significantly higher than that of the Pearson dimension reduction method.It can effectively reduce the dimension of the pepper canopy spectrum and improve the accuracy of the model.Among the four models based on Pearson-GAPLS dimension reduction,RF model had higher accuracy than PLSR,BP and LSSVM models,with R²,RMSE and RPD of 0.80,3.50 and 2.21,respectively.The model can be well applied to the estimation of chlorophyll content in pepper leaves.（2）The inversion model of SPAD value for different varieties of pepper was constructed.From the correlation analysis and model construction based on SPAD values and near-ground hyperspectral data of four pepper varieties,it was found that there were differences in canopy spectral characteristics of different pepper varieties.The prediction accuracy of SPAD values of modeling set,validation set and full sample of pepper leaves was higher in line pepper than in Chaotian pepper.Except CI_{red edge},the correlation coefficients of vegetation indices based on reciprocal logarithmic spectral optimization are higher than those of classical vegetation indices.The random forest model based on log1/R-VI has good accuracy in both modeling set and validation set,which is suitable for the estimation of SPAD value of different varieties of pepper.（3）The SPAD value inversion model based on UAV hyperspectral data is constructed.Based on the correlation analysis of the selected pepper sample leaf SPAD value and the UAV hyperspectral data and the constructed model discovery:The SPAD value of pepper leaves is inversely proportional to the reflectivity.The sensitive bands of capsicum chlorophyll in UAV hyperspectral are mainly concentrated in 400nm～700nm.The correlation between the spectrum after first-order differential treatment and SPAD value is the best.The first-order differential spectrum at 671nm wavelength is negatively correlated with chlorophyll content,and the correlation coefficient is-0.69.The model based on reciprocal logarithmic spectrum generally has high accuracy.The GA-LSSVM model based on differential spectroscopy performed best,with R²,RMSE and RPD values of 0.84,1.41 and 2.24,respectively,followed by the RF model based on the reciprocal logarithm spectrum,with R²,RMSE and RPD values of 0.83,1.57 and 2.13,respectively.（4）The inversion of SPAD value of pepper leaves based on space-ground spectral parameters in the study area was realized,and a high universal inversion model was selected.From the optimal preprocessing method and optimal model based on UAV hyperspectral and near-ground hyperspectral,the effects of various transformation spectra on the model are different.In the near-earth spectrum,the RF model built with the reciprocal logarithmic spectrum has the highest accuracy.In the UAV spectrum,the model accuracy of the reciprocal logarithm spectrum is generally high.By comparing and analyzing the accuracy of different models,it can be seen that the RF model can be used as a prediction model for SPAD value,and its R²,RMSE and RPD values are 0.83,1.57 and 2.13,respectively.In summary,the RF model based on the reciprocal logarithmic spectrum can predict the SPAD values of different data sources and different varieties of pepper leaves,which is universal.